A STANDARDS-BASED GOVERNANCE FRAMEWORK FOR ARTIFICIAL INTELLIGENCE ASSISTED SYSTEMATIC LITERATURE REVIEWS UNDER THE EU JOINT CLINICAL ASSESSMENT
Author(s)
Aditi Bajpai, PharmD, MHS1, Angeline Babitha Dhas, BS1, Maria Rizzo, BSc, MS2, Lingamaiah Doddolla, B.Pharm, M.Pharm3, Revanth M, B.E.3, Viji Queen V, Pharm.D, PGDMW4, Meghan Oates-Zalesky, MSc5.
1MadeAi, Cambridge, MA, USA, 2Independent, Kent, United Kingdom, 3MadeAi, Nagercoil, India, 4MadeAI, Nagercoil, India, 5Chief Marketing Officer, MadeAi, Cambridge, MA, USA.
1MadeAi, Cambridge, MA, USA, 2Independent, Kent, United Kingdom, 3MadeAi, Nagercoil, India, 4MadeAI, Nagercoil, India, 5Chief Marketing Officer, MadeAi, Cambridge, MA, USA.
OBJECTIVES: Artificial intelligence (AI) increasingly supports evidence generation and systematic literature reviews (SLRs), yet no standardized approach exists for AI use in the Health Technology Assessment Regulation (HTAR) Joint Clinical Assessment (JCA). We aimed to develop a standards-based governance framework for AI-assisted SLRs spanning methodology, evidence synthesis, and accountable AI use, anchored to best-practice standards to support credible JCA submissions.
METHODS: We mapped four governance principles: human-in-the-lead control, audit trail, reproducibility, and provenance to established best-practice standards from the three areas an AI-assisted SLR for JCA must satisfy. For SLR methodology (searching, screening, reporting), we applied the EUnetHTA information-retrieval guideline, PRESS, and PRISMA-S within PRISMA 2020; for responsible AI use, we applied ISPOR ELEVATE-GenAI as the primary reporting standard, with TRIPOD+AI for prediction models and the NICE position statement on AI; and for JCA requirements, the HTACG dossier guidance. These standards were encoded as configurable checks with automatic deviation flags.
RESULTS: The identified best-practice guidance was synthesized into a standards-based governance framework for AI-assisted SLRs wherein each workflow stage—from PICO specification and SLR protocol development through literature searching, study selection, and data extraction—was aligned with established standards, paired with human oversight, and supported by audit trails and provenance. The framework generated an AI-use audit report documenting AI-assisted steps and human sign-off, alongside a search documentation package aligned with best-practice reporting standards. EU HTA methodological requirements were operationalized as configurable checks that flagged deviations from predefined methodological requirements.
CONCLUSIONS: The proposed standards-based governance framework provides a transparent, traceable approach for AI-assisted SLRs under the EU HTA Regulation by aligning workflow outputs with established methodological, reporting, and AI-governance standards. This enables AI-assisted activities to be justified against recognized best practices while maintaining human oversight and auditability. Future work should evaluate the framework through application to an EU JCA dossier.
METHODS: We mapped four governance principles: human-in-the-lead control, audit trail, reproducibility, and provenance to established best-practice standards from the three areas an AI-assisted SLR for JCA must satisfy. For SLR methodology (searching, screening, reporting), we applied the EUnetHTA information-retrieval guideline, PRESS, and PRISMA-S within PRISMA 2020; for responsible AI use, we applied ISPOR ELEVATE-GenAI as the primary reporting standard, with TRIPOD+AI for prediction models and the NICE position statement on AI; and for JCA requirements, the HTACG dossier guidance. These standards were encoded as configurable checks with automatic deviation flags.
RESULTS: The identified best-practice guidance was synthesized into a standards-based governance framework for AI-assisted SLRs wherein each workflow stage—from PICO specification and SLR protocol development through literature searching, study selection, and data extraction—was aligned with established standards, paired with human oversight, and supported by audit trails and provenance. The framework generated an AI-use audit report documenting AI-assisted steps and human sign-off, alongside a search documentation package aligned with best-practice reporting standards. EU HTA methodological requirements were operationalized as configurable checks that flagged deviations from predefined methodological requirements.
CONCLUSIONS: The proposed standards-based governance framework provides a transparent, traceable approach for AI-assisted SLRs under the EU HTA Regulation by aligning workflow outputs with established methodological, reporting, and AI-governance standards. This enables AI-assisted activities to be justified against recognized best practices while maintaining human oversight and auditability. Future work should evaluate the framework through application to an EU JCA dossier.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
MSR172
Topic
Health Policy & Regulatory, Methodological & Statistical Research, Study Approaches
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
No Additional Disease & Conditions/Specialized Treatment Areas